# IMPACT is Working

In Launching the Innovation Renaissance I wrote:

…teacher pay in the United States seems more like something from Soviet-era Russia than 21st century America. Wages for teachers are low, egalitarian and not based on performance. We pay phys ed teachers about the same as math teachers despite the fact that math teachers have greater opportunities elsewhere in the economy. As a result, we have lots of excellent phys ed teachers but not nearly enough excellent math teachers….

Soviet style pay practices helped to eventually collapse the Soviet system and the same thing is happening in American education. Michelle Rhee is no longer the DC Chancellor but IMPACT, the teacher evaluation system developed under her tenure, is in place. IMPACT uses student scores to evaluate teachers but also five yearly in-class evaluations, three from the school administrator and two from master educators from outside the school. Evaluations are meant not only to reward but also to discover and spread best teaching practices.

The results from IMPACT are starting to come in and they indicate that pay for performance is encouraging low quality teachers to leave, good quality teachers to get better, and high quality teachers to continue teaching and improve even further.

Perhaps not surprisingly the schools with the poorest students see the most teachers leave and they also see the largest gains in student performance as average teacher quality rises. From a new NBER paper by Adnot et al.:

More than 90 percent of the turnover of low-performing teachers occurs in high-poverty schools, where the proportion of exiting teachers who are low-performers is twice as high as in low-poverty schools.

…Our estimates indicate that there are consistently large gains from the exit of low-performing teachers in high-poverty schools. In math, teacher quality improves by 1.3 standard deviations and student achievement by 20 percent of a standard deviation; in reading these figures are 1 standard deviation of teacher quality and 14 percent of standard deviation of student achievement.

These are big effects especially when multiplied over many generations of students.

Hat tip: Eric Crampton at Offsetting Behaviour.